An ANN-based framework for estimating inconsistency in lateral placement of heterogeneous traffic

Abstract

The lateral placement is the transverse position of a moving vehicle across the carriageway width. Studies on lateral placement have gained importance over time specifically for two reasons; (1) lateral placement data has become a crucial input in most of the traffic flow simulation models, and (2) detection of wheel positions helps in determining the riding quality and the distressed portions on a pavement surface. In the case of heterogeneous traffic flow and loosely enforced lane discipline, the estimation of lateral placement of vehicles deals with additional complexity. In such cases, vehicles take any lateral position left empty by other surrounding vehicles while moving in a mixed traffic stream. This results in an inconsistent lateral trajectory of vehicles. Further, this inconsistency is primarily governed by the subject vehicle type and other prevailing factors of the traffic stream. On this background, the present study forwards a Neural Network-based approach to quantify the inconsistency associated with the lateral placements chosen by different vehicle categories in a mixed traffic situation. In addition, a sensitivity analysis revealed how the inconsistency in lateral placement varies suggestively with the change in traffic and road geometric factors.